Description
The talk examines power noise modelling through Gaussian Processes for secure True Random Number Generators.
While revisiting one-sided fractional Brownian motion, we obtain novel contributions by quantifying posterior uncertainty in exact analytical form, establishing quasi-stationary properties, and developing rigorous time-frequency analysis. These results are applied to model oscillator fluctuations of power-noise type, enabling closed-form entropy expressions for TRNGs and a novel GPU-accelerated simulation technique valuable for studying non-standard post-processing.
This work bridges machine learning techniques and signal processing to solve hardware security applications.
Keywords
Gaussian Process, Power Noise, True Random Number Generator, Fractional Brownian Motion, Entropy Estimation, Hardware Security, GPU Acceleration
Practical infos
Next sessions
-
Attacking the Supersingular Isogeny Problem: From the Delfs–Galbraith algorithm to oriented graphs
Speaker : Arthur Herlédan Le Merdy - COSIC, KU Leuven
The threat of quantum computers motivates the introduction of new hard problems for cryptography.One promising candidate is the Isogeny problem: given two elliptic curves, compute a “nice’’ map between them, called an isogeny.In this talk, we study classical attacks on this problem, specialised to supersingular elliptic curves, on which the security of current isogeny-based cryptography relies. In[…]-
Cryptography
-